Real-time characterization of intermediates in the pathway to open complex formation by Escherichia coli RNA polymerase at the T7A1 promoter

Abstract

We have used time-resolved x-ray-generated hydroxyl radical footprinting to directly characterize, at single-nucleotide resolution, several intermediates in the pathway to open complex formation by Escherichia coli RNA polymerase on the T7A1 promoter at 37°C. Three sets of intermediates, corresponding to two major conformational changes, are resolved as a function of time; multiple conformations equilibrate amongst each other before the next large structural change. Analysis of these data in the context of published structural models indicates that initial recognition involves interaction of the UP element with the α-subunit C-terminal domain and binding of the σ subunit to the -35 sequence. In the subsequent isomerization step, two complexes with footprints extending into the -10 region can be differentiated as the DNA becomes distorted during nucleation of strand separation. During the final isomerization step, the downstream double helix becomes embedded in the β/β′ jaws, leading to a transcriptionally active complex.

De novo RNA synthesis during DNA transcription is a highly regulated cellular process carried out by RNA polymerase. RNA polymerase binds to DNA and diffuses one-dimensionally to the promoter region (1-4); it then must make a number of interactions within the promoter as a prerequisite to forming a relatively stable complex. This complex is then in a state that favors isomerization, leading to strand separation from approximately -12 to +3 with respect to the site of transcription initiation (5). Each of these steps is a possible site for regulation of the transcription levels. Extensive studies have been conducted on the structure of RNA polymerase-DNA complexes at equilibrium resulting in the characterization of the interaction of each of the enzyme's subunits with specific promoter elements (6, 7). However a direct characterization of the interactions formed at each step of the recognition process is still lacking.

The T7A1 promoter used for the studies described here is one of the strongest known prokaryotic promoters (8). Its -35 sequence, TTGACT, is very close to the consensus (TTGACA), and its -10 sequence, GATACT, deviates from consensus (TATAAT). In addition, this promoter contains, from -42 to -80, a sequence rich in adenine and thymine residues, also known as an UP element (9-11). On other promoters this sequence has been shown to both increase the rate of promoter binding and stimulate isomerization to the open complex (12-15). Kinetic studies on the formation of an RNA polymerase-promoter complex revealed the presence of a sequence of isomerization steps leading to the formation of an open complex (16-18). Furthermore, by decreasing the isomerization rates at lower temperatures, one or more intermediates in the pathway from the initial closed complex to the final open complex were isolated and characterized (19-22). However, the large, and sometimes nonlinear, temperature dependence of some of the steps in this pathway (16, 18, 23) and the occurrence of intermediates at low temperatures that were not detected on the normal kinetic pathway (21) suggest that the structure of intermediates and/or the mechanism of promoter recognition and DNA opening could differ at low temperatures compared with normal physiological conditions.

We have carried out time-resolved hydroxyl radical footprinting experiments (24) at a physiological temperature, 37°C, during T7A1 promoter recognition by Escherichia coli σ⁷⁰ RNA polymerase. Analysis of our results allows a direct structural and kinetic characterization of the intermediates in the pathway of open complex formation by RNA polymerase on promoter DNA at a single base resolution.

Methods

Synthesis of the DNA Fragment Containing the T7A1 Promoter. First, 5′-Alexa Fluor 647-labeled 172 base pairs of DNA containing promoter A1 was prepared by PCR using DNA primers A1-up(-91) for the nontemplate strand (NTS) and A1-down(+81) for the template strand (TS). Either one of the primers was labeled via 5′-aminolink with Alexa Fluor 647 fluorescent dye. The DNA fragments were purified by HPLC on a monoQ column by using a MiLiChrom chromatograph (EcoNova, Novosibirsk, Russia).

RNA Polymerase Purification. RNA polymerase was isolated from E. coli RL721 (provided by Robert Landick, University of Wisconsin, Madison) according to Kashlev et al. (25) with a final purification on heparin-Sepharose to resolve the holo (α₂, β, β′, ω, and σ) and core (α₂, β, β′, and ω) forms of the enzyme. The protein preparation used was 100% saturated with σ as measured by quantification of the bands on a SDS gel and was 100% active in promoter binding as measured by gel-shift assay.

Beamline Characteristics and Stopped-Flow Machine. These experiments were carried out at the beamline ID10A at the European Synchrotron Radiation Facility (Grenoble, France). Exposure times between 0.5 and 2 ms, depending on beam current and reaction conditions, resulted in cleavage of DNA within a single-hits kinetic regime. When using these exposure times, there is a 20-30% quenching of Alexa Fluor 647 fluorescence (data not shown). The stopped-flow apparatus we designed for these experiments used two separate motors and a sample collector at the end of the machine to sequentially mix and expose 11 samples. The mixing was carried out with a mixing T of our own construction, and the dead time for mixing was 50 ms. All of the tubings of the stopped flow were immersed in a water bath controlling the sample temperature (±0.5°C) (more details on the experimental setup can be found in Supporting Methods, which is published as supporting information on the PNAS web site).

Binding Reactions. The binding reactions were performed by mixing 10 μl of 100 nM Alexa Fluor 647-DNA in BBc20 buffer (10 mM cacodylate, pH 7.5/20 mM NaCl/6 mM MgCl₂) and 10 μl of 200 nM RNA polymerase at 37°C inside the stopped-flow machine for each time point. The RNA polymerase was dialyzed immediately before the experiment against BBc20 buffer on a floating membrane VSWP02500 (Millipore) to remove glycerol that would quench the hydroxyl radicals. The irradiated samples (100 μl) were mixed with carrier DNA (0.02 mg/ml) and precipitated with 3 vol of ethanol. The DNA fragments produced from hydroxyl radical cleavage were resolved on 8% denaturing PAGE by using an ALF Express II DNA analyzer (Amersham Pharmacia Biotech).

Data Analysis

Quantification and Normalization of Time-Resolved Footprints. The output file of an electrophoresis run on the ALF Express DNA analyzer II (fluorescence intensity vs. time for each lane on the gel, Fig. 1 A and B) was converted into ASCII format. By using the software igor (WaveMetrics, Lake Oswego, OR), a linear baseline was subtracted from each data set. The Peak Fitting Module of the software originpro (OriginLab, Northampton, MA) was used to fit the peaks in each lane to a Lorenzian curve. Approximately 130 peaks were fit simultaneously, covering the whole region interacting with the polymerase. The values for peak areas within each lane were divided by the average value of the area of several peaks within the same lane that were not protected by the polymerase in the course of the binding reaction. Normalized area values then were divided by the values for the corresponding peaks in the lane of DNA cleaved in the absence of protein; the resulting values were then relative change compared with naked DNA, Φ (Fig. 1C; see also Fig. 6, which is published as supporting information on the PNAS web site) (26).

Fig. 1.

Time-resolved footprinting profiles. (A and B) Footprint profiles of the T7A1 promoter fragment labeled with Alexa Fluor 647 at the 5′ end of the NTS (A) and TS (B) as a function of mixing time with RNA polymerase. Each line graph corresponds to one lane on the gel. Each peak is a fragment 1 nt longer than the fragment from the previous peak. The mixing time for each lane is shown next to the line graph. (C) The bar plots show the ratio of the area of each peak relative to that in the naked DNA (lane at 0 sin A and B), Φ, for the two strands (nontemplate at the top and template at the bottom), at the end of the binding reaction. The bar at nucleotide 0 is missing because in the conventional numbering of the nucleotides in a promoter 0 is not used. See Fig. 6 for additional graphs of intermediate time points.

Fit of Kinetic Data to Single and Double Exponential Equations. The progression curves (Φ vs. time) of appearance of protection from hydroxyl radical cleavage at each nucleotide position were fit individually to single or double exponential expressions as follows:

[1]

and

[2]

where A and B are the signal amplitudes and k, k_A and k_B are the observed, apparent rate constants. The upper (U) and lower (L) limits from these fits were used to normalize the data from 0 to 1, where 0 corresponds to the value in the absence of polymerase and 1 corresponds to the value for the area of the peak at its minimum, resulting in the value of the fractional saturation at that site. We observed that often, within the same region, the difference in the value of the rate and amplitude measured for different nucleotides was within the error derived from the fit. These curves for nucleotides within the same region, and from two independent experiments, were combined and then fit again to reduce the error. To determine whether an expression containing a single or a double exponential better described the results in each data set, a visual analysis of the residuals (Fig. 2 A) was carried out, as was an F test (see Table 1, which is published as supporting information on the PNAS web site).

Fig. 2.

Kinetics of the change in fractional saturation in different regions of the promoter as a function of incubation time of RNA polymerase with the T7A1 promoter fragment. Each section shows the change in fractional saturation (1-Φ normalized from 0 to 1) for multiple neighboring nucleotides from two independent experiments. The nucleotide positions are designated with “m” (minus) for nucleotides upstream and “p” (plus) for those downstream of the site of the start of transcription (designated as +1). (A)(Top) Comparison of fits to a single (dashed curve) and double (continuous curve) exponential equation for the data set m9-m12 TS. (Middle and Bottom) Shown are the residuals (the difference between the data points and the fit curve) for the two fits. The residuals for the double-exponential fit (Middle) are evenly distributed on either side of the fit curve, whereas the residuals for the single-exponential fit (Bottom) show a periodic pattern of distribution. From this analysis, and from an F test comparison of the fits (see Supporting Methods), we conclude that the double-exponential equation results in a better fit of the data. This analysis was carried out for each data set (data not shown). (B) Comparison of a double-exponential fit performed on the whole set of data points (dashed curve) or excluding the zero point (continuous curve) for the data set m33-m31 TS. The fit in the absence of the zero time point results in a more even distribution of the residuals and in the determination of the amplitude of the burst. The fit also was carried out for the other data sets where the fit in the absence of the zero point resulted in a better fit (m55-m54 TS and m43-m45 TS). (C) Representative plots of the kinetics of protection for specific sites on the TS. (D) Plots of the kinetics of protection for sites on the NTS. In the plots for the upstream sites, a decrease in fractional saturation can be observed at the longer time points; in this case, the fit was carried out in the absence of the points in the decreasing portion of the curve.

Simulation of the Kinetic Model and Global Fitting of the Data. Simulation and global fitting of the data by using a kinetic model and parameter sensitivity analysis was carried out with the software berkeley madonna (R. I. Macey and G. F. Oster, 2001, Univ. of California, Berkeley, CA). A kinetic model was built starting from the simplest possible pathway (P + R ⇔ A, where P is the promoter and R is the RNA polymerase) then gradually adding more intermediates until a simulation of the progress curves from the results obtained on the TS could be obtained by using as initial estimates the rates determined by previous studies (23, 27, 28) and those from the fits of the binding curves. A model with three simple sequential steps (P + R ⇔ A ⇔ B ⇔ C) could describe some of the binding curves but could not account for the differences in amplitudes and the smaller differences in rates as a function of nucleotide position. To do the latter, each step was associated with the formation of more than one complex, each having a footprint of a different size (Figs. 3C and 4A; see below and Supporting Methods for additional details). The resulting models were used to simultaneously fit the data from the TS (Fig. 4B).

Fig. 3.

Summary plot of the amplitudes and the apparent rates of appearance of the protections. (A) Apparent rates of appearance of protection as a function of nucleotide position. The black bars correspond to the protections on the NTS, whereas the gray bars correspond to those protections on the TS. The error bars correspond to the joint confidence interval at the 65% confidence limit of the fits shown in Fig. 2. Values of the slowest rate (≈0.02 s^-1) are not shown; also omitted are those values from nucleotides whose kinetic curves were too noisy due to a low signal amplitude (-53, -56, -20 to -17 TS). (B) The amplitude of each kinetic phase is shown as the percentage of the total amplitude as a function of the nucleotide position on the promoter. The burst phase amplitudes are colored white, the amplitudes corresponding to the faster rate measured are colored black, and those corresponding to the slow phase are gray. The amplitude of the faster phase of the p3,p5,p7 data set when it is fit to a double exponential is marked with a black horizontal line; values for the fit to the simplest model, a single exponential, are shown in Table 2. The known RNA polymerase domains responsible for a given protection are shown next to their corresponding signal (6, 7). (C) Schematic representation of the regions of the TS protected in different intermediates shown in the mechanism in Figs. 4 and 5.

Fig. 4.

Possible kinetic models for open complex formation. (A) Possible kinetic models for the binding of RNA polymerase to the T7A1 promoter. Values for representative microscopic rate constants resulting from the global fit are shown; the units are s^-1 except for the forward rate constant for the formation of A, which has the units of M^-1·s^-1.(B) Result of a global fit of data sets for sites on the TS by using a linear kinetic model. Each point in the plot is the average of the values shown in the plots in Fig. 2 A-C.

Results and Discussion

Time-Resolved X-Ray Induced OH-Footprinting of RNA Polymerase on Promoter DNA. An example of the gel profiles obtained with DNA labeled by Alexa Fluor 647 at the 5′ terminus of either strand is presented in Fig. 1 A and B. The profile in the first row from the top represents the cleavage pattern of the DNA by hydroxyl radicals in the absence of RNA polymerase. The peak intensity corresponds to the cleavage efficiency at each nucleotide. As the incubation time with RNA polymerase is increased, the intensity of peaks in specific regions decreases. Fig. 1C represents the profiles of the peak area measured by peak fitting (see Methods) relative to that in naked DNA (Φ) at the end of the binding reaction (126.5 s).

Kinetics of RNA Polymerase Binding Show Multiphasic Behavior. At the earliest time point, 180 ms, a weak pattern of modulated protection is present from position -62 to -17 on the NTS and from -65 to -15 on the TS (Figs. 1 and 6). In the region from -65 to -21 TS, the amplitude of the protection appearing within the dead time of the experiment is sufficiently large to observe in the kinetic plots (Fig. 2B). The kinetics of the subsequent increase of protection can be described by either a single- or double-exponential equation (Eqs. 1 and 2) depending on the position of the nucleotides on the promoter (Fig. 2). The protection of nucleotides from +9 to +17 NTS and from +3 to +20 TS is monophasic and appears at a rate similar to that of the slow phase present in the sites further upstream, which instead follow a biphasic behavior (Fig. 3; see also Table 2, which is published as supporting information on the PNAS web site).

On the NTS, a decrease in the extent of protection is observed in the upstream region at the longer incubation times; the amplitude of this effect being greatest in the most upstream sites, gradually decreasing until a slow increase in protection was observed from -12 to +20 NTS (Fig. 2D). This decrease makes it impossible to determine the real extent of fractional saturation at these sites. The data for the upstream sites (from -61 to -17 NTS) were fit to a single exponential only up to the point where the protection begins to decrease. As a result, those rates measured from fits at these sites do not reflect the true apparent rates of formation of the corresponding complex but only provide an estimate of their magnitude. The appearance of protection of the bases at -11 and -12 and from -5 to +8 NTS is biphasic with the fast phase accounting for 50% of the total amplitude.

Amplitude of Fast Phases Changes As a Function of Nucleotide Position. In the case where the appearance of protection is best described by a double-exponential equation, a difference in the amplitude of the two phases as a function of nucleotide position is observed. This result is shown, for example, by comparing the plot for nucleotides -23 to -21 TS with that of nucleotides -8 to -5 TS (Fig. 2C). The amplitude of the burst also changes with respect to nucleotide position (Fig. 2 B and C). The sum of the amplitudes of the burst and of the fast phase decreases from 80% to 20% from upstream to downstream on the TS (summarized in Fig. 3B).

Identification of the Pattern of Protection of Intermediates in the Pathway. The difference in the rates with which nucleotides become protected as a function of their position on the promoter shows that there are at least three steps in the formation of the final complex. The first step results in the appearance of protection within the dead time of the experiment (<180 ms); this event is the “burst” observed in the plots in Fig. 2 B and C and shown in the summary in Fig. 3B; the second step results in protection appearing at a slower rate (k_obs ≈ 1 s^-1), corresponding to the fast phase observed in the kinetic curves. Protection arising during the third step appears at a rate that is an order of magnitude slower (k_obs ≈ 0.02 s^-1) (Table 2).

To account for the discernibly different amplitudes of the burst at positions -55 to -54, -45 to -43, -33 to -31, and -23 to -21 TS, we propose that the first binding step leads to the formation of three different complexes, each resulting in the protection of a region of different size (A, A′, and A″) (Fig. 3C). The subsequent step, by a similar logic based on differences in signal amplitude and observed rates of the fast phase, leads to two complexes with footprints of different size, one with an additional protection from -14 to -9 TS and a larger one from -14 to +3 TS (B and B′, respectively). The final step, resulting in the appearance of protection down to +20, corresponds to the formation of complex C. These complexes constitute the minimal number of intermediates necessary to account for these results.

A specific pattern of protection for the A intermediates on the NTS cannot be confidently assigned because of the smaller amplitude of the signal compared with the noise in the data. Here we propose that, on this strand, the protection up to -6 (where the slow phase of protection increase becomes apparent) corresponds to B, the protection from -5 up to +8 corresponds to B′ (the protections in this region have a slower apparent rate for the fast phase), and the protection up to +17 corresponds to C.

Global Fit to Different Kinetic Models. To determine the possible relationship between the intermediates, we have built different kinetic models and used them to carry out simulations of the kinetics and global fits of the data sets. As described above, the real extent of fractional saturation at certain sites on the NTS was difficult to ascertain. Data for the TS therefore were used for the modeling and global fitting. This set of kinetic curves can be fit by models in which the structures of A and B intermediates are in equilibrium with each other before the next conformational change (Fig. 5). As a result of this degeneracy, only a fraction of the population can isomerize to the next intermediate conformational states (as seen by a decrease in the amplitude of the faster kinetic phase). The data thus can be fit equally well by either a linear or branched pathway model (such as those shown in Fig. 4A), resulting in different sets of values for the microscopic rate constants for the different steps in the pathway (a plot of the fit is shown in Fig. 4B). The relatively slow isomerization rates observed at T7A1 and other promoters containing nonconsensus sequences are possibly due to the limited amount of the intermediate that is in the correct conformation to isomerize to the next step. Here, we suggest that during promoter recognition, RNA polymerase conducts a search in a funnel-like conformational space, in a manner similar to that proposed to take place during protein and RNA folding (29, 30).

Fig. 5.

Model of proposed conformational changes occurring during promoter recognition and open complex formation by RNA polymerase. The structure of the polymerase is that determined by the laboratory of S. Darst (7). RNA polymerase subunits β and β′ are in two different shades of green, whereas the α N-terminal domains are in light and dark blue. The α-CTD, absent in the crystal structure, have been added as blue spheres. The σ subunit is in red. A simplified cartoon of the DNA is drawn to scale. The TS is in dark purple and the NTS is light purple. The contacts shown between the enzyme and the DNA reflect the decreased solvent accessibility of nucleotides in specific positions on the promoter during the formation of the complex as determined in this work by time-resolved hydroxyl radical footprinting.

Structural Description of Real-Time Intermediates on the Pathway to Open Complex Formation. Because of their small size, hydroxyl radicals cleave the DNA backbone at all solvent accessible sites (31). Each pattern of protection thus corresponds to a specific structure of the complex (Fig. 5). By using the available published data from crystallographic and extensive biochemical studies describing RNA polymerase contacts with different regions of the promoter (6, 7, 32), we can assign those subdomains of different subunits of the protein that interact with DNA during the recognition steps at this promoter (Figs. 3B and 5).

Complexes A, A′, and A″. The C-terminal domain of the α-subunit (α-CTD) can bind to AT-rich sequences (11, 12, 33). Several AT-stretches are found between positions -40 and -80 on the T7A1 promoter. We therefore attribute the protection at nucleotides -43 to -45 and -53 to -56 on the TS and -46 to -50 and -58 to -60 on the NTS to an interaction with the two α-CTDs. These two regions constitute the proximal and distal binding sites for the α-CTDs. Protection from -33 to -31 TS and -41 to -39 NTS is due to region 4.2 of σ (7, 34-36). On the TS, we observed the appearance of protection at -56 to -53, -46 to -42, and -33 to -31 within the first second of the reaction. This result supports recent evidence of a specific interaction between α-CTD and region 4.2 of σ in the stabilization of initial binding (37-39). Here, we directly establish that the interaction of these two subunits with their respective binding sites occurs at the initial steps of promoter recognition.

The protected region from -23 to -21 TS and -25 to -29 NTS on the promoter comes into close contact with the β′ zinc-binding domain (6, 7). The protection at -23 to -21 appears with a burst that is too small to be measured within experimental error, however, the amplitude of the fast phase is larger than that measured for protected regions downstream (see also the m21-m23 data set in Fig. 4B). We therefore assign the protection extending down to -21 to an intermediate (A″) that is in rapid equilibrium with A and A′.

Complexes B and B′. These intermediates are represented by five stretches of modulated protection separated by ≈10 bp. This pattern of protection is due to the RNA polymerase sitting on one face of the double helix. In complex B (Fig. 3), an exception to this 10-bp periodicity in the protection pattern is found in the -10 region, where protection appears from -15 to -6 NTS in addition to -14 to -9 TS and, to a lesser extent, -20 to -17 (TS) (Fig. 6). Most of these protections are due to σ regions 3.0, 2.4, and 2.3 protruding above the surface of the polymerase (7). However, in the models of the closed complex, where the double helix continues on a straight path from σ region 2, the nucleotides from -8 to -6 NTS are not near any polymerase subunits (40). These nucleotides can only become protected once the DNA has been bent toward the opening of the β/β′ channel; such a bend could result from an increased flexibility of the DNA due to nucleation of opening (18, 40, 41). The loss of periodicity at -10 is not observed in the footprint pattern of a closed complex formed at 4°C but is observed in the complex formed at 20°C (20). We propose that an initial interaction of σ regions 3.0, 2.3, and 2.4 between -20 and -6 NTS and -14 to -9 TS causes a (temperature-dependent) kink in the DNA (41), placing the two strands in position for further strand separation.

In the B′ intermediate (Fig. 3), the protection extends to +8 NTS and to +2 TS. Only the NTS is protected from +3 to +8, suggesting that in this intermediate the double helix has not yet entered the β/β′ channel but that the NTS is already in close proximity to its surface. The difference in protection between B and B′ agrees with the model proposed by Gralla and coworkers (42, 43) for nucleation and opening of the double helix. Time-resolved fluorescence studies of tryptophan residues in RNA polymerase (23) revealed a nonlinear temperature dependence of the isomerization rate that could be explained by the presence of the structural isoforms B and B′ identified here.

Positions -11 and -12 NTS become protected in a biphasic fashion, unlike adjoining nucleotides down as far as -6 that show monophasic kinetics (Fig. 2B), whereas nucleotides at -10 and -9 do not become protected at all. This result suggests that the protection of these key nucleotides for nucleation of DNA opening (-11 and -12) occurs in the step after the interaction with nucleotides -8 to -6.

Complex C. The rate of this step is similar to that measured for the appearance of an active complex by using the abortive initiation assay (27, 44). Because the protection on the TS from +3 to +20 only appears in this last phase, it is possible that this strand is not correctly placed within the active site until the downstream double-stranded region is bound in the “jaws” of β and β′. During the last step, a partial loss of protection upstream on the NTS is observed in the binding sites of α-CTDs and σ region 4 and, to a lesser extent, down to -20 (Fig. 2B). A similar change was observed in this region by Li and McClure (22) using DNase I footprinting of the closed complex of the λPRmup-1 Δ265 promoter. This decrease in protection could result from a loss of contacts due to a twisting of the double helix arising from strand separation. This alleviation of upstream contacts could facilitate escape from the initiating complex. A too-large stability conferred by some of the upstream contacts is deleterious to the formation of an active transcription complex, and the presence of an UP element can slow down promoter escape and in some cases has been shown to inhibit promoter clearance (13, 45).

A salient feature of this pathway is that the α-CTDs direct the polymerase to the promoter by binding to the UP element, which, by reason of its particular structural features of narrow minor groove and overall bent double helix, acts as the primary recognition site (33, 36, 37). The correct binding of σ region 4.2 to the -35 sequence is partly dependent on the flexibility of the DNA at the TG step at -35 and at the -39 tetrad GTAT, facilitating the α-CTD-σ region 4.2 interaction (37). This kind of indirect readout of the DNA, based more on its potential structure than on direct sequence recognition, is likely to play an important role in the initial steps of binding at this promoter. Further studies of DNA and RNA polymerase mutants will establish whether a given structure can be stabilized by optimized interactions, thus increasing its accumulation and the rate of the following isomerization rate. This stabilization also could be caused by the presence of transcription factors or alternate DNA structures such as supercoiling.